- Boost-users - lists.boost.org

[signals] thread-safe post-disconnect() cleanup
by Frank Mori Hess 26 Feb '07

26 Feb '07

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 We could implement a thread-safe post-disconnect callback by using the existing tracking functionality. That is, the connection body could hold a "cleanup" shared_ptr with a custom deleter. The custom deleter would run the user's cleanup callback. A weak_ptr copy of the cleanup shared_ptr would be kept just like it was any other tracked object. When disconnect() is called, the cleanup shared_ptr would be reset. If the signal is being invoked concurrently when disconnect() is called, the cleanup code won't get run until the last invocation is done with the slot and destroys its copy of the cleanup shared_ptr. BTW, I've added support for a two-parameter version of signals::track() to thread_safe_signals/track.hpp. It seems to work okay, the only thing I don't like about it is it implies there a coupling required between the value and the tracked shared_ptr, when really there isn't. - -- Frank -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.1 (GNU/Linux) iD8DBQFF3iTv5vihyNWuA4URAkzWAKDHL0MjcFGmAvKbWoBYmhNN5RiCMACgypCS SLJPrYcUaR0u4YTTS35lEZQ= =dCrK -----END PGP SIGNATURE-----

5 58

Re: [Boost-users] [boost::any] - reference vs pointer
by Benjamin Winkler 26 Feb '07

26 Feb '07

References can not be copied. Thus, you can't make boost::any contain a reference (which is what mean when you write boost::any<double&>, I assume). In your example, msgargs.push_back(dref) calls the any-constructor any(const T&) with T being double, and this copies the value, not the reference itself. When you call msgargs.push_back(dptr), it calls the any(const T&) with T being double*. The value is copied again, only this time, the value is a pointer. Benjamin Winkler >Hi all, > >I've been trying out boost::any in a messaging system to package up function arguments. A message >call is basically > >Message::Call(std::vector<boost::any>& args) >{ >try >...a few any_cast<..> >...fire the actual function... >catch >...etc >} > >It's all good, except for one niggling problem. Say I have a double-precision value that I need >changed inside the message, so it needs to be passed as a double* or a double& in an argument. The >problem is, if I create a double& inside a boost::any, it doesn't point to the right memory location >once I'm inside the message, and it looks like it's copying the original dereferenced value and >giving me a reference to the copy. If I create a double* inside a boost::any, I get a new pointer >inside the message like the reference, but because the pointer target is correct everything works. >The code looks like this: > >... >... >std::vector<boost::any> args; >double d = 1.5; >double* dptr = &d; >msgargs.push_back(dptr);//works fine... >double& dref = d; >msgargs.push_back(dref);//seems to copy d instead of a reference to d... >... >msg->Call(args); >... > >I can live with this, I just have to pass "large" objects by pointer so they don't get copied. BUT: > Can anyone tell me why references are creating a copy of the original and not a copy of the >reference, or how I could avoid the copying? Or am I doing something dumb? > >Damien >_______________________________________________ >Boost-users mailing list >Boost-users(a)lists.boost.org >http://lists.boost.org/mailman/listinfo.cgi/boost-users >

3 2

[FileSystem] Getting modified date and time from a file?
by Lawrence Spector 26 Feb '07

26 Feb '07

Does boost provide a portable way to do this? I didn't see an obvious way; I was hoping others might have ideas. Thanks in advance, Lawrence

5 5

[program_options] Documentation of configuration file format?
by Tarjei Knapstad 26 Feb '07

26 Feb '07

I can't seem to find any documentation on what kind of file formats are accepted when calling boost::program_options::parse_config_file(...) Am is missing it or is it missing? :) -- Tarjei

4 7

BoostCon'07 Registration Now Open
by info＠boostcon.com 26 Feb '07

26 Feb '07

The following message is a courtesy copy of an article that has been posted to comp.lang.c++.moderated,comp.std.c++,comp.lang.c++ as well. I'm pleased to announce that the official website for the inaugural Boost conference is up and registration is open: http://www.boostcon.com Due to the very high quality of the submissions we've received, our program committee has been able to assemble a lineup of speakers and topics the likes of which I haven't seen anywhere (http://www.boostcon.com/program) Combine that with the relaxed atmosphere of the conference venue (http://www.boostcon.com/location) and I think we can all look forward to a truly unique and memorable week of working and learning together. I invite you to visit http://www.boostcon.com/registration to register. -- Dave Abrahams Boost Consulting www.boost-consulting.com

1 0

Filesystem library questions
by Manish Motwani 26 Feb '07

26 Feb '07

Hi, We're creating a new cross-platform dispersed storage technology. Backing up a file from one platform and restoring to another (or the same) is a requirement. We have found boost to be very useful for threads, date/time, program options, etc. We're also currently using the boost filesystem libarary, but we can't figure out why many of the file-metadata query and modifier functions are missing. Can someone from boost let us know if there's any way of querying complete file-metadata (including read/write/execute permissions on windows and linux, ACLs, creation-time in windows, etc.). I found only the following function in version 1.32 (the one we're using): last_write_time (read and write modification time). symbolic_link_exists (check if a file is a sym link). And these new functions have been added since: create_hard_link create_symlink .. and some functionality to get the type of file is also there in the trunk version. But what about the rest of metadata filesystems (POSIX, NTFS, FAT32, HFS, etc) provide? Is this ever going to be included in the library? Is there a way to access this information any other way? I suspect these questions may have been asked by others, so can someone point me to where I can find code to complement the boost filesystem library to handle file-metadata? Thanks, Manish

1 0

Boost date time and MinGW
by Franck Guillaud 25 Feb '07

25 Feb '07

Hello World, I'm doing some experiments to use boost (1.33.1) under Win XP and with MinGW (5.1.3) environment. Boost was compiled with bjam with no particular options. I'm facing some problems on linking with boost_date_time. Here is the simple test program: #include <iostream> #include <string> #include <boost/date_time/posix_time/posix_time.hpp> using namespace std; using namespace boost::posix_time; int main (int argc, char** argv) { char dummy; ptime d1(min_date_time); ptime d2(max_date_time); cout << "min:" << to_simple_string(d1) << endl; cout << "max:" << to_simple_string(d2) << endl; cin.get(dummy); return 0; } Compiling lines are : g++ -DUSE_DATE_TIME_PRE_1_33_FACET_IO -I"E:\dev\boost_1_33_1" -O0 -g3 -Wall -c -fmessage-length=0 -MMD -MP -MF"main.d" -MT"main.d" -o"main.o" "../main.cpp" g++ -L"E:\dev\boost_1_33_1\stage\lib" -o"TestBOOST.exe" ./main.o -lboost_date_time-mgw-mt-d-1_33_1 I get the error (simplified output) : E:/dev/boost_1_33_1/boost/date_time/gregorian/greg_weekday.hpp: undefined reference to `boost::gregorian::greg_month::as_long_string() const' If I remove the lines with *to_simple_string*, I get no error. From the docs [*] it seems there may be a problem for streams support for some compilers, so I tryed to define USE_DATE_TIME_PRE_1_33_FACET_IO in my project, but with no effect. Does it needs to be done when compiling boost ? Is there any other way to solve my link problem ? (appart not using string output for dates :-) ) Thanks for any help. Franck,e- [*] http://www.boost.org/doc/html/date_time/details.html#date_time.buildinfo

2 2

Any reasons why an exception would not be caught by BOOST_CHECK_THROW?
by Stephen Torri 25 Feb '07

25 Feb '07

I have an exception class that is not getting caught my BOOST_CHECK_THROW. Here is a similar setup to what I have: class Base_Exception : public virtual std::exception { public: virtual ~Base_Exception() throw{} virtual const char* what () const throw() = 0; virtual const char* name () const throw() = 0; boost::uint32_t const& id () const throw(); private: boost::uint32_t id; }; class Child_Exception : public virtual Base_Exception { public: enum ERROR_ID { INVALID_API_INPUT, CANNOT_FIND_TARGET, INTERNAL_LIBRARY_ERROR, }; Child_Exception ( ERROR_ID message_id ); virtual ~Child_Exception () throw () {} virtual const char* what () const throw(); virtual const char* name () const throw(); private: static const char* m_exception_name; static const char* m_messages[]; }; I find that when I use the BOOST_CHECK_THROW ( expression, Child_Exception ) that I still get the test crashing with an uncaught exception. Here is the actual output: terminate called after throwing an instance of 'libreverse::errors::API_Exception' what(): Internal library error has occurred. At this point consider the library unreliable. unknown location(0): fatal error in "test_wrong_sources": signal: SIGABRT (application abort requested) *** errors detected in test suite "Architecture Type Detector suite"; see standard output for details FAIL: test_arch_type_detector I can send the whole test file but I hope this is enough to understand the nature of the problem. Is there a FAQ about BOOST_CHECK_THROW? Stephen

1 0

[graph] birected -> directed adapter?
by Jens Müller 25 Feb '07

25 Feb '07

http://www.boost.org/libs/graph/doc/reverse_graph.html reverses edge directions. Is there an adapter (i.e., has someone written such a thing) that combines in and out edges in a bidirected graph and gives a view as if the graph were undirected? I'd need this for doing BFS independent of edge directions in a bidirected graph.

1 0

Bug in options_description.cpp when offer a long program options to boost::program_options
by Jack Shi 25 Feb '07

25 Feb '07

If you provide a long program options description while using boost, you will encounter such issue, if you using debug version, you will see a debug assertion failed alert, if you using release version, your program will be crashed. following codes can be used to reproduce this issue #include <iostream> #include <boost/program_options.hpp> using namespace std; int main(int argc, char* argv[]) { namespace po = boost::program_options; po::options_description desc("supported options"); desc.add_options() ("help,h", "display this help message,this help will be printed out if you not provide option here!"); std::cout << desc << "\n"; } I have trace this issue, it is located in function format_paragraph in the file options_description.cpp, namespace { void format_paragraph(std::ostream& os, std::string par, unsigned first_column_width, unsigned line_length) { ...... // prevent chopped words // if (lastchar != ' ') && // ((exists(lastchar + 1) && (lastchar + 1 != ' ')) if ((*(line_end - 1) != ' ') && ((line_end < par_end) && (*line_end != ' '))) { // find last ' ' in the second half of the current paragraph line string::const_iterator last_space = find(reverse_iterator<string::const_iterator>(line_end - 1), reverse_iterator<string::const_iterator>(line_begin - 1), ' ') .base(); if (last_space != line_begin - 1) { ...... } if option description is long enough, the find function will be called, and "reverse_iterator<string::const_iterator>(line_begin - 1)" will lead to this issue, may "line_begin - 1" access a invalid space. If there have some guys to confirm it was a bug in boost?

2 1